Responsible Microservices
summary
Concrete rules for when microservices pay off: split parts that show multiple rates of change (recommendation or search engines) and use git-based churn heatmaps to find hotspots; migrate via a data-driven Strangler proxy that logs inputs/outputs and runs new and legacy paths in parallel; design independent lifecycles and pipelines so services can release daily; monitor with the four golden signals and use circuit breakers for failure isolation. Small, replaceable services (rewrite in two weeks) win.
AI-written summary · about
chapters (55)
- 0:00Intro
- 1:53Forming two pizza teams.
- 3:19We have to consider the cost of the added complexity.
- 4:45If parts of your system evolve at different speeds...
- 5:16Maybe the Inventory system is really stable.
- 7:18Start with your source code management tool...
- 8:44Software archeology time!
- 9:39Chad Fowler created Turbulence based on churn vs. complexity.
- 11:31Who is working on what?
- 13:25Last commit around the Super Blue Blood Moon Eclipse?
- 14:25Strangler Application.
- 14:35Reduces the risk of a big bang cutover.
- 15:23We can go a step further and apply a data driven approach.
- 16:19Put a proxy layer between the client and the legacy system.
- 16:36Drives test cases for the new functionality.
- 16:57You can run the new in parallel with the old.
- 17:27Don't be surprised if the old system is wrong!
- 19:10Run experiments. A/B testing.
- 20:01Speed matters.
- 20:48Monoliths often have dictionary sized getting started guides.
- 22:06It can take months for a new developer to get up to speed.
- 22:16What was your longest stretch to get to productive team member?
- 22:51Goodbye 80 hour manual regression suites.
- 23:03Forget the one-off performance test.
- 23:13Use the right tools for the job!
- 23:26Each microservice can use the mix of tests that make sense.
- 23:36We can practice hypothesis drive development
- 24:00How do you know?
- 25:29Focus on "paved roads."
- 27:08Deploy early, deploy often.
- 27:40Need to develop trust in the process.
- 27:50Spring Cloud Pipelines.
- 29:45For example - how much capacity will you need?
- 30:37It was in our best interest to over allocate resources.
- 30:52Gave us single digit resource utilization.
- 31:41Annual budgets make it difficult to add capacity smoothly.
- 32:14Odds are the order processing system has a unique scaling needs.
- 32:34Monitoring to the rescue!
- 33:06Four Golden Signals.
- 33:39Spring Boot Actuator!
- 35:08Often use bailing twine and duct tape...
- 35:42Look for failure points.
- 36:04When month end falls on the Super Blue Blood Moon.
- 36:53It is hard. We are really good at thinking through the happy path.
- 37:29The circuit breaker pattern.
- 37:41Redirects to the fallback mechanism.
- 38:22Chaos engineering.
- 39:01Teams develop deep expertise.
- 39:11Simplifies the hiring and training processes.
- 39:40Ops can focus on the primary environment.
- 41:27We aren't forced down the square peg round hole path.
- 41:38It's great right? Each team can use just the right tool for the job!
- 42:41Pick from these 3 flavors. Won't work for you? Let's talk.
- 44:14With great power comes great responsibility.
- 44:25Avoid the temptation of resume driven design.
